Please wait...
Old hardware is just not worth messing with anymore after a certain point.
Upgrading from a Ryzen 3900X to an i7-14700KF not because I really wanted to, but because all the offboard USB ports on the motherboard (except for most of the back-panel ones) are now completely fried.
Normally I try to at least wait for the next process node before upgrading--these two are both "7nm". But the current Intel generation is a dud, so if building a new system, this seems to be where the smart(ish) money is. If I was going to have to replace the motherboard and PSU, I figure just build somethiing new, even if the CPU and RAM were salvageable...
Parts will arrive today, so hopefully I can get back to having a fully-working system soon.
the last couple have been.
Fri Mar 28 2025 15:07:45 UTC from LoanShark
. But the current Intel generation is a dud,
I think Raptor Lake is OK now that they've ironed out the microcode issues. A little less power efficient than Ryzen 9000 series at the top end of the curve, but more efficient at idle which is where it counts. Better buy at current market prices from what I can tell.
Upgrading from a Ryzen 3900X to an i7-14700KF not because I really
wanted to, but because all the offboard USB ports on the motherboard
(except for most of the back-panel ones) are now completely fried.
Ouch. Took a pretty bad hit?
At that point I'd be worried that none of the surviving parts would ever be quite right again. Starting a new build (except for maybe the storage or whatever) seems the right way to go, especially if it's a daily driver that you rely on.
Good luck with your build!
Ouch. Took a pretty bad hit?
At that point I'd be worried that none of the surviving parts would
ever be quite right again. Starting a new build (except for maybe the
storage or whatever) seems the right way to go, especially if it's a
daily driver that you rely on.
I'm not entirely sure what happened. The wireless card I had installed was pretty typical: an ASUS with a Mediatek chipset, where the WiFi domain connects via PCIe, and the Bluetooth domain talks to one of your unused internal USB2 headers on the motherboard. As far as I can tell, basically all the products on the market work this way.
It had been working fine under Windows for quite some time, but I started trying to use it under Linux, realized I had to upgrade `linux-firmware` to get it working, and went through the whole procedure of upgrading Ubuntu etc. It initially seemed flakey when I finally got it working *at all*, but then seemed to be working just fine for an hour or two (not sure if that was anything I did), and then it decided to fry the motherboard's USB connector that it was hooked up to. I messed around a bit, tried all the other internal USB2 headers, bought another WiFi card, etc, no dice, decided to just upgrade the system.
My thinking is the same as yours; many components are suspect, so I only moved the storage devices over. Things had been a bit wonky for a while, a few times Linux caused the NVMe to crash. The PSU was a suspect, and it's rather ancient so I condemned it as well.
The new build seems mostly fine (I'm typing this from it right now) but there's still a bit of wonkiness where both Linux and Windows logs show occasional correctable PCIe bus errors on the port that talks to the Samsung 990 Pro (which is fairly new, and I've got enough experience with similar issues on prior Samsung products that I suspect they may eventually push a firmware fix.)
Oh, I meant to add that all the other front-panel USB ports (and one of the back-panel ones) got fried quite some time ago by devices that were probably shorting (e.g. a bad Xbox controller)
So it's been a case of gradually losing USB ports for a few years now, and just saying "I'll live with it until it becomes unacceptable", and it finally became unacceptable.
I had a USB attached litecoin miner once. It caught fire. Literally, flames coming out of the fan. Thankfully i only ran it when i was home so i was 2 feet away when it did.
It took out ALL the USB ports on the PC it was connected to.
Decades ago, had a PS catch fire on a mini-tower ( 386 days ). Took out the mother board, CPU, RAM even the HD and CDROM drive was toast. Never seen such a thing before. Sure, bits and pieces but not *everything*
Wait, no, it might have been a 486.. or perhaps even a early Pentium. its been so long i dont remember now. 20 years or so, and it wasn't new at the time..
A few comments on the Raptor Lake setup -
* Asus motherboard, the TUF B760-Plus Wifi I think it's called. BIOS defaults to fairly conservative PL1/PL2 settings that cost me about 17% performance in CineBench
* I'm using fairly high-end cooling, albeit air cooling. You can lift the power limits in the BIOS. There's also a middle-of-the-road option that lifts the power limits BUT limits the CPU temperature to 90C. This is what I've settled on. It makes about a 40W difference in power consumption and about a 2.5% difference in CineBench, which might be within the margin of error. Power limits are user configurable to whatever wattages you feel are right for your setup.
This, plus Intel's recent root cause analyis I think should address any concerns about CPUs burning themselves up. Unless you really just choose not to believe Intel's root cause
Decades ago, had a PS catch fire on a mini-tower ( 386 days ).
Back when I worked at a VAR doing PC builds and troubleshooting, every time we got a Packard Bell in we knew the PSU was likely dead. I worked on one that you could smell it burning up.
Packard Smell.
I also don't recommend plugging in Xbox controllers that may or may not have been covered in cat vomit, although that may be a topic for Pussy>
Back in the late 80s i got a call about a monitor that was not coming on. Since it was a tube, i had the guy flip the power as i looked inside the vents in the back, just to see if it was getting power at all. Flame about hit me in the face. One of the resistors decided to cut loose. "ok, ill get you a new monitor, hang on" lol I guess if i was able to smell, i might have noticed it when i entered the room.
Had a dell desktop do that too. About caught the wall on fire behind it. Next day a laptop caught on fire, also a dell, due to its poor design. Never been fond of them since.
Another time, same company as the flaming resistor, different area, one of the engineers that designed automotive chips called. "my monitor isn't turning on". In this office the monitors faced away from the front door so i could see the cord was missing instantly. I walked in, went over to his machine, picked the power cord off the floor and plugged it in to the monitor, and didn't say a word... He watched me do it.. turned bright red.. "oh god. im sorry "
Or the time i forgot to discharge the tube on a terminal, as i was busy talking to the user, and stuck my paw in it to swap out the board..
Ah fun times. Kids today will never understand
Sat Mar 29 2025 21:53:45 UTC from LoanSharkDecades ago, had a PS catch fire on a mini-tower ( 386 days ).
Back when I worked at a VAR doing PC builds and troubleshooting, every time we got a Packard Bell in we knew the PSU was likely dead. I worked on one that you could smell it burning up.
Packard Smell.
It might take more time before I really know for sure, but so far, the PCIe bus errors from the NVMe seem to go away if I limit the link speed to PCIe v3.0. Not sure how I feel about that... I paid for v4.0 obviously.
Don't want the hassle of going through a warranty claim and transferring all my data to a new drive, either, though.
To me, it sounds like its struggling to stay alive. Slowing it down is just a short term bandaid and it will hard fail next.
Been there too many times with SSD. Suddenly starts slowing down trying to keep up with error corrections. Then dead. ( a couple of times before i could get the stuff off. it died during that process )
Mon Mar 31 2025 13:23:33 UTC from LoanShark
It might take more time before I really know for sure, but so far, the PCIe bus errors from the NVMe seem to go away if I limit the link speed to PCIe v3.0. Not sure how I feel about that... I paid for v4.0 obviously.
Don't want the hassle of going through a warranty claim and transferring all my data to a new drive, either, though.
I guess with older tech too.. Floppies, spinny drives, tape.. whatever.
But most of the time back then for me it was just 'poof' .. rarely got a chance to get stuff off.
Mon Mar 31 2025 13:41:23 UTC from Nurb432
Been there too many times with SSD.
Been there too many times with SSD. Suddenly starts slowing
down trying to keep up with error corrections. Then dead. ( a
Ooops, text client screwup. Anyway, this device is supposedly fine at the block layer (99% health remaining based on total bytes written) so this is a PCIe PHY issue. not sure what to make of it.
Basically, it's hard to determine how widespread this is as a percentage of the Samsung 990 Pro user base; forum posts don't give me the denominator. But there's enough complaints about ASPM issues and PCIe bus errors to suspect that there's some sort of unacknowledged issue and that this may be fairly "normal" for this device, i.e. I might not be able to run it at full speed without "correctable" bus errors and/or might not be able to run it with ASPM fully enabled without the device occasionally falling off the bus. Some people posting on forums have gone so far as to RMA their entire datacenter fleet of 990 Pro's and move to a different vendor.
I had a few incidents of the device entirely ceasing to respond (dropping off the PCI bus, as if it wasn't even connected until a power cycle restored it), but so far that has only happened under Linux and only in the previous machine and only prior to a BIOS update...
You can thank Microsoft for that.
Tue Apr 01 2025 13:44:29 UTC from LoanShark
only happened under Linux